Using the Right Benchmarks
"Benchmark data" refers to control data against which stop data can be compared to determine if any racial or ethnic group is being stopped at a disproportionate rate. The right benchmark can provide the racial and ethnic demographic for any given locality, whether it be an urban intersection or a state highway. Stop data can then be compared to the demographic, and a statistical analysis can be conducted which will help determine if some racial groups are being stopped more frequently than their demographic presence.

We strongly believe that collecting the right benchmark, or understanding the true demographic of a locality, is essential to procuring valid results on assessing profiling activities. If the assumed demographic is suspect, then the comparison to stop data may yield invalid results – the analysis will be meaningless.

The only way to determine the true demographic for any given locality is to survey for racial and ethnic traffic – to conduct Point to Point Surveys. This means that the racial and ethnic mix of individuals traveling through a locality must be identified and recorded. A schedule must be developed to survey carefully chosen locations according to a randomly selected time schedule. If the right locations are surveyed according to the right schedule, then the demographic for a given locality may be assumed.

Other benchmarks, such as census data on population demographics, will not serve as reliable benchmarks. Census data measures static populations, that is, the geographic demographic of households. Highway and pedestrian traffic represent transient populations. People work in different locations from which they live, and travel in different routes and different ways to get there. Additionally, tourism, business trips, and other populations not measured in census data such as university populations make the comparison suspect. For example, in New Jersey v. Soto (1996) and Wilkins v. Maryland State Police (1996), it was found that census data did not accurately predict highway transient traffic.

Certainly, data sources such as census, arrest, search, and accidents can provide insight into these complex problems. To date, none of these “benchmarks” alone or in combination have been able to accurately assess traffic data. These data sources simply do not accurately reflect driving populations, and will not yield valid results. It is an axiom of statistical analysis that inaccurate data will yield misleading results.